How much VRAM does Gemma 4 E4B need?
Gemma 4 E4B needs 4.5 GB for its weights at 4-bit, 5.7 GB with an 8K context and 6.1 GB with 32K. At 8-bit, 9.1 GB.
- Parameters
- 8B
- Active per token
- 4.5B
- Longest context
- 128K
- Released
- Mar 2026
- License
- Apache 2.0
- Strongest at
- Chat, Coding, Docs
Memory at each precision
| Precision | Weights | With 8K | With 32K | With 128K |
|---|---|---|---|---|
| 16-bit | 14.9 GB | 16.1 GB | 16.4 GB | 17.9 GB |
| 8-bit | 7.9 GB | 9.1 GB | 9.5 GB | 11.0 GB |
| 6-bit | 6.1 GB | 7.3 GB | 7.7 GB | 9.2 GB |
| 5-bit | 5.3 GB | 6.5 GB | 6.8 GB | 8.3 GB |
| 4-bit | 4.5 GB | 5.7 GB | 6.1 GB | 7.6 GB |
| 3-bit | 3.6 GB | 4.8 GB | 5.2 GB | 6.7 GB |
| 2-bit | 2.8 GB | 4.0 GB | 4.3 GB | 5.8 GB |
| 1.6-bit | 1.5 GB | 2.7 GB | 3.1 GB | 4.6 GB |
One conversation, 16-bit context cache, runtime overhead included.
What runs it
| Machine | Best fit | Context | Speed | First reply |
|---|---|---|---|---|
| RTX 3060 12 GB | 8-bit | 128K | 43 tok/s | 0.7 s |
| RTX 4060 Ti 16 GB | 8-bit | 128K | 35 tok/s | 0.4 s |
| RTX 3090 | 8-bit | 128K | 103 tok/s | 0.2 s |
| RTX 4090 | 8-bit | 128K | 110 tok/s | 0.1 s |
| RTX 5090 | 8-bit | 128K | 177 tok/s | 0.1 s |
| 2× RTX 3090 | 8-bit | 128K | 103 tok/s | 0.1 s |
| RTX PRO 6000 Blackwell | 8-bit | 128K | 177 tok/s | 0.1 s |
| M4 Max 128 GB | 8-bit | 128K | 53 tok/s | 1.4 s |
| M5 Max 128 GB | 8-bit | 128K | 58 tok/s | 0.4 s |
| Ryzen AI Max+ 395 128 GB | 8-bit | 128K | 26 tok/s | 0.8 s |
| DGX Spark 128 GB | 8-bit | 128K | 33 tok/s | 0.3 s |
| M3 Ultra 256 GB | 8-bit | 128K | 73 tok/s | 0.8 s |
| M5 Ultra 512 GB | 8-bit | 128K | 96 tok/s | 0.2 s |
| H100 SXM | 8-bit | 128K | 277 tok/s | 0.1 s |
| DGX Station GB300 | 8-bit | 128K | 422 tok/s | 0.1 s |
| 8× H200 NVL | 8-bit | 128K | 345 tok/s | 0.1 s |
Best fit is the highest quality, up to 8-bit, that leaves 16K of context. First reply is the wait for a 2K prompt. Speeds are rough, plus or minus 25%.
How to run it
- Ollama
ollama run gemma4:e4b- llama.cpp
llama-server -hf unsloth/gemma-4-E4B-it-GGUF- vLLM
vllm serve google/gemma-4-E4B-it- MLX (Mac)
mlx_lm.chat --model mlx-community/gemma-4-e4b-it-4bit
Questions
How much VRAM does Gemma 4 E4B need?
Gemma 4 E4B needs 4.5 GB for its weights at 4-bit, 5.7 GB with an 8K context and 6.1 GB with 32K. At 8-bit, 9.1 GB.
Can Gemma 4 E4B run on an RTX 4090?
Yes. At 8-bit it holds 128K of context and writes about 110 tokens a second.